In a 2 node V2 cluster, with a DRBD resource(heartbeat resource/ not
OFC) and file system resource on top of it there appears to be a need
for a delay in failovers where the other node died.

 

I can fail over manually, by putting the active node in standby and
everything seems fine

 

Failing over by cutting power to the active node seems to result in a
bad file system mount.

 

A startup delay on the file system resource and on the application seems
to resolve the issue.

 

I was hoping to gain more confidence through better understanding, can
anyone explain the reason this occurs?

 

Is it that the heartbeat DRBD resource takes longer to failover when the
other node is dead than when it is alive?

 

Should I need the application delay, or is the file system delay
sufficient?

 

In other words, once DRBD has done its thing, can the file system be
mounted and accessed in a very short time frame, or does the mount
requires a delay as well?

 

Should the DRBD resource not wait for completion before the next ordered
resource is allowed to start?

 

Thanks

 

Damon Estep

 

 

_______________________________________________
Linux-HA mailing list
[email protected]
http://lists.linux-ha.org/mailman/listinfo/linux-ha
See also: http://linux-ha.org/ReportingProblems

Reply via email to